Compare commits

...

1 Commits

Author SHA1 Message Date
Dai Ha 428a12af62 fleetd #722: reconcile presence that arrives before a session's registry entry
CI / shell-tests (pull_request) Failing after 9s
CI / contract (pull_request) Successful in 50s
CI / build (pull_request) Failing after 1m48s
CI / shell-tests (push) Failing after 9s
CI / contract (push) Successful in 47s
CI / build (push) Failing after 1m46s
A member whose MCP contact lands between launcher.spawn() and registry.put()
had its presence marked, but the SPAWNING -> READY transition that markPresent
triggers found no registry entry yet and silently did nothing. The mark then
persisted while registration left the session in SPAWNING, with nothing to
retry the transition. That left the session undeliverable to reclaim/seat
accounting even though it was present and deliverable.

Add SessionManager.reconcilePresence, called right after registry.put in both
the plain-spawn and worktree-spawn paths, to retry the transition for a
terminal already marked present. One private helper serves both call sites.

Tests cover both orderings (contact-then-register and register-then-contact)
for both spawn paths, plus a terminal never marked present staying in
SPAWNING. The contact-then-register tests use a new PresenceRacingLauncher
test double that marks presence from inside spawn(), before acquire()'s own
registry.put runs.
2026-10-04 19:23:02 +02:00
4 changed files with 195 additions and 0 deletions
@@ -255,6 +255,10 @@ public final class SessionManager implements TurnListener {
handle.id(), handle.terminalId(), resolvedProfile, actualRole, cwd, ownerTerminal, now, now, 0,
MemberSession.State.SPAWNING, null, null, handle.charterReceipt(), handle.agentSessionId());
registry.put(handle.id(), session);
// A presence contact that already arrived for this terminal found no registry
// entry to transition and gave up silently. Retry it now that one exists; remove
// this call and such a session stays in SPAWNING even though it is present.
reconcilePresence(handle.terminalId());
handles.put(handle.id(), handle);
log.debug("acquired session id={} terminal={} profile={} owner={}",
handle.id(), handle.terminalId(), session.profile(), session.ownerTerminal());
@@ -809,6 +813,10 @@ public final class SessionManager implements TurnListener {
handle.charterReceipt(),
handle.agentSessionId());
registry.put(handle.id(), session);
// A presence contact that already arrived for this terminal found no registry entry to
// transition and gave up silently. Retry it now that one exists; remove this call and
// such a session stays in SPAWNING even though it is present.
reconcilePresence(handle.terminalId());
handles.put(handle.id(), handle);
log.debug("acquired worktree session id={} terminal={} profile={} branch={} path={}",
handle.id(), handle.terminalId(), session.profile(), session.branch(), session.worktree());
@@ -983,6 +991,20 @@ public final class SessionManager implements TurnListener {
transitionByTerminal(terminalId, MemberSession.State.SPAWNING, MemberSession.State.READY);
}
/**
* Completes a newly registered session's {@code SPAWNING -> READY} transition when {@code
* terminalId} was already marked present before this ran. A terminal never marked present is
* left in {@code SPAWNING}; it reaches {@code READY} normally through {@link #onReady} once
* its own contact arrives. Callers must run this only once the session's registry entry is
* already visible — {@link #onReady}'s transition matches against that entry, and reconciling
* before the entry exists finds nothing to transition.
*/
private void reconcilePresence(String terminalId) {
if (terminalId != null && !terminalId.isBlank() && presence.isPresent(terminalId)) {
onReady(terminalId);
}
}
/**
* Lifecycle hook: a message was delivered into the worker — it is now busy on a turn.
* The turn count is bumped and the activity timestamp is refreshed. A {@code DONE} session
@@ -0,0 +1,92 @@
package dev.ltms.fleet.session;
import dev.ltms.fleet.inject.MemberPresence;
import dev.ltms.fleet.peer.Capability;
import dev.ltms.fleet.peer.PeerHandle;
import dev.ltms.fleet.peer.PeerLauncher;
import dev.ltms.fleet.peer.SpawnRequest;
import dev.ltms.fleet.placement.PlacementDecision;
import java.util.List;
import java.util.Set;
/**
* {@link PeerLauncher} decorator that marks presence for a spawned terminal before returning its
* handle to the caller — the contact-then-register ordering fleetd #722 covers, where the
* terminal's MCP contact lands before {@link SessionManager#acquire} runs its own
* {@code registry.put}. The presence view is set after construction, via {@link #presence},
* because it is owned by the {@link SessionManager} this launcher is passed into.
*/
final class PresenceRacingLauncher implements PeerLauncher {
private final PeerLauncher delegate;
volatile MemberPresence presence;
PresenceRacingLauncher(PeerLauncher delegate) {
this.delegate = delegate;
}
@Override
public PeerHandle spawn(SpawnRequest req) {
PeerHandle handle = delegate.spawn(req);
presence.markPresent(handle.terminalId());
return handle;
}
@Override
public PeerHandle spawn(SpawnRequest req, PlacementDecision decision) {
PeerHandle handle = delegate.spawn(req, decision);
presence.markPresent(handle.terminalId());
return handle;
}
@Override
public Set<Capability> capabilities() {
return delegate.capabilities();
}
@Override
public Set<Capability> capabilitiesFor(String profileName) {
return delegate.capabilitiesFor(profileName);
}
@Override
public Set<String> profiles() {
return delegate.profiles();
}
@Override
public String defaultProfile() {
return delegate.defaultProfile();
}
@Override
public String effectiveCwd(SpawnRequest req) {
return delegate.effectiveCwd(req);
}
@Override
public List<String> parityOverlay(String profileName) {
return delegate.parityOverlay(profileName);
}
@Override
public List<?> list() {
return delegate.list();
}
@Override
public int reapOrphanWorkers() {
return delegate.reapOrphanWorkers();
}
@Override
public void stop(String id) {
delegate.stop(id);
}
@Override
public boolean clearContext(String id) {
return delegate.clearContext(id);
}
}
@@ -422,6 +422,53 @@ class SessionManagerTest {
"turn completion moves BUSY → DONE");
}
// --- fleetd #722: registration and presence must reach READY whichever lands first --------
@Test
void registerThenContactReachesReadyForPlainSpawn() {
FakeHerdr herdr = new FakeHerdr();
SessionManager sessions = sessionManager(herdr);
MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
sessions.asPresence().markPresent(session.terminalId());
assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
"a presence contact that arrives after registration reaches READY");
}
@Test
void contactThenRegisterStillReachesReadyForPlainSpawn() {
// The racing launcher marks presence for the spawned terminal from inside spawn() —
// before SessionManager.acquire's own registry.put runs — modeling an MCP contact that
// lands in that window.
FakeHerdr herdr = new FakeHerdr();
FleetConfig.Profile cfg = new FleetConfig.Profile(
"ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN",
List.of("ccs", "ltms-local"), "tab", "fleetd-workers",
"worker: {profile} #{n}", null, null, null);
ClaudeCodeLauncher workers = new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null);
PresenceRacingLauncher race = new PresenceRacingLauncher(workers);
SessionManager sessions = new SessionManager(race);
race.presence = sessions.asPresence();
MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
"a presence contact that lands before registry.put must still reach READY");
}
@Test
void aTerminalNeverMarkedPresentStaysSpawningAfterRegistration() {
FakeHerdr herdr = new FakeHerdr();
SessionManager sessions = sessionManager(herdr);
MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
assertEquals(MemberSession.State.SPAWNING, sessions.get(session.paneId()).orElseThrow().state(),
"registration alone must not advance a terminal that was never marked present");
}
@Test
void releaseTearsDownWorkerAndRemovesFromRosterAndIsIdempotent() {
FakeHerdr herdr = new FakeHerdr();
@@ -159,6 +159,40 @@ class WorktreeSessionManagerTest {
assertEquals(expectedPath, s.cwd(), "session cwd is the worktree path");
}
// --- fleetd #722: registration and presence must reach READY whichever lands first --------
@Test
void registerThenContactReachesReadyForWorktreeSpawn() {
FakeHerdr herdr = new FakeHerdr();
FakeWorktrees worktrees = new FakeWorktrees().withRepoRoot("/repo").withPrefix("/wt");
SessionManager sessions = new SessionManager(workerService(herdr), worktrees);
MemberSession session = sessions.acquire("ltms-local", null, "/caller/proj", "term_primary",
new WorktreeRequest("cb-722", null));
sessions.asPresence().markPresent(session.terminalId());
assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
"a presence contact that arrives after worktree registration reaches READY");
}
@Test
void contactThenRegisterStillReachesReadyForWorktreeSpawn() {
// The racing launcher marks presence for the spawned terminal from inside spawn() —
// before SessionManager.acquireWithWorktree's own registry.put runs — modeling an MCP
// contact that lands in that window.
FakeHerdr herdr = new FakeHerdr();
FakeWorktrees worktrees = new FakeWorktrees().withRepoRoot("/repo").withPrefix("/wt");
PresenceRacingLauncher race = new PresenceRacingLauncher(workerService(herdr));
SessionManager sessions = new SessionManager(race, worktrees);
race.presence = sessions.asPresence();
MemberSession session = sessions.acquire("ltms-local", null, "/caller/proj", "term_primary",
new WorktreeRequest("cb-722", null));
assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
"a presence contact that lands before worktree registration must still reach READY");
}
@Test
void worktreeArchitectAcquireAlsoBindsItsSlot() {
FakeHerdr herdr = new FakeHerdr();